Process-Tracing Methods: Foundations and Guidelines

Chapter 5. Causal Inference and Process-Tracing Methods

68 Chapter 5 Causal Inference and Process-Tracing Methods Social scientists use theories in the same manner that we use maps—to simplify an immensely complex reality. Yet whereas cartographers engage in descriptive inference when they make maps, social scientists are also interested in going a step beyond describing what happens to study causal relationships, explaining how and why social phenomena occur. Theories of causal mechanisms are, for example, simplifications of reality that predict what causal forces are important in explaining the occurrence of a phenomenon . Yet social scientists are not interested in theories as purely hypothetical thought experiments. Theories need to be empirically validated to make sure that they accurately represent reality. This chapter deals with the interrelationship between theory and empirical reality. After a theory is developed, we need to empirically validate its accuracy. We do so by testing whether empirical evidence strengthens or weakens our confidence in the validity of the theory as an explanation of a phenomenon. Given the difficulty of measuring complex social phenomena such as democracy, we cannot fully measure what is happening in reality. Even with the best measuring instruments, we can gain only a selected sample of observations of a given phenomenon. Therefore, we are forced to infer from a small set of empirical observations that a theory was the cause of the phenomenon . In other words, we make an inferential leap from what we can observe empirically to conclude that an underlying causal explanation exists. What enables us to jump from a set of empirical observations to infer whether a causal mechanism exists in a specific case in process-tracing research ? Is the logic that enables inferences to be made the same in process- Causal Inference and Process-Tracing Methods 69 tracing as the frequentist logic of inference that underlies the methodological prescriptions for qualitative case study research as suggested by King, Keohane, and Verba (KKV)(1994)? In process-tracing case studies, we attempt to analyze whether a theorized causal mechanism exists in an individual case. Therefore, we are interested in making what can be termed within-case inferences, meaning that we use empirical evidence collected from a particular case to infer that all of the parts of a hypothesized causal mechanism were actually present in that case. Process-tracing methods cannot be used to make cross-case inferences that involve concluding based on evidence drawn from a sample of comparable cases that a causal relationship exists across a population of a given theoretical phenomenon. Other forms of inferential tools, such as comparative methods, are necessary to make cross-case inferences. In this chapter, we discuss why mainstream inferential tools used, for example, in classical statistical analysis and comparative cross-case methods cannot be used to make within-case inferences about causal mechanisms. Here, we continue the argument that our methodology must be brought in line with ontology (Hall 2003). In particular, we illustrate that given that the ontology of causality used in process-tracing differs from that of both classical statistical analysis and comparative methods, the inferential tools used in these methods do not apply in process-tracing. This discussion is followed by the presentation of the Bayesian logic of inference as a solution to the problem of making within-case inferences in process-tracing. The chapter concludes by discussing in more detail the types of inferences that can be made using process-tracing methods and, equally important, what types of inferences cannot be made. 5.1.The Type of Inferences Made in Process-Tracing Process-tracing methods are used when we want to investigate whether causal mechanisms are present or absent in an in-depth case study (theorytesting ), when we want to build a theorized causal mechanism based on an in-depth examination of the evidence in a single case (theory-building), or when we want to explain a particular outcome, such as why Nazi Germany was not effectively deterred from aggression against Poland by threats from France and the United Kingdom (explaining-outcome). In all three variants of process-tracing, we are investigating a causal mechanism (or mechanisms) in what is effectively a single-case study. An important distinction across the variants is whether the ambition is to gener- 70 Process-Tracing Methods alize beyond the single case (theory-centric) or is more focused on accounting for the outcome in the single case (case-centric). In a theory-centric single-case study, the researcher is interested...